YouTube Announces New AI-Generated Music Feature. Enter the age of indistinguishability between the person and the AI.
【YouTube Announces New Feature to Make Music with “AI-Generated Voice of Pop Stars”】
・YouTube to launch new feature to make music with AI-generated voices of famous pop stars, YouTube announced on 16th
・Experimental AI feature called “Dream Tracks” will allow “selected U.S. creators” to generate 30-second tracks for YouTube “shorts,” according to YouTube
・Starting on the 16th, music can be generated by Dream Tracks by 9 artists including Alec Benjamin, Charlie Puth, Charlie XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan
・All elements – lyrics, instruments, and voices – are generated by AI tools, YouTube said, adding that the software is powered by Lyria, “the most advanced music generation model to date” from Google’s DeepMind.
・YouTube has introduced measures requiring creators to disclose any content they create or fabricate that appears realistic, including the use of AI tools.Other platforms, such as TikTok, have also introduced tools and requirements to clearly flag deepfake and AI content as fake or falsified
・As more realistic content from generative AI models like ChatGPT, which generates text and images for OpenAII, becomes more prevalent, there is a growing demand for tools to help people distinguish between real and fabricated content. Leading AI companies have promised to add “watermarks” to AI-generated content to help people determine whether content online is real or fake
・DeppMind states that audio published through its “Lilia” will be “watermarked” with “content that is inaudible to the human ear and does not detract from the listening experience” so that it can be detected after the audio content has been manipulated.
These are the quotes from the article
We are entering an era in which it is impossible to tell whether it is the person himself/herself or an AI. How will the distinction be made?
YouTube has announced a new feature. It seems that AI can generate not only famous artists’ voices, but also lyrics and instruments in that artist’s style. Below is what was created using the feature.
It is already indistinguishable from the real thing.
It seems that this feature is still only available to a limited number of 9 creators (artists), but it will eventually become available to the general public.
Come to think of it, fake videos of politicians have been in the news lately.
What makes President Biden speak Japanese is truly amazing. Japanese with the voice exactly as it is!
Video Borrowed from→【If you use AI to make President Biden speak Japanese, here’s what happens: “He’s so fluent, he doesn’t sound out of place,” “He even sounds like him!】
With generative AI, it is easy to create something like this.
We can tell that this is a fake, though, because there is no way that President Biden speaks Japanese,
However, this video clearly shows that we live in an era in which we can make as many fake videos as we want, as if they were real (the person himself). The same goes for the music mentioned above.
As noted in the article above,
The major AI companies say that anything created by a generative AI will have a marker (watermark) that will tell you that it was created by a generative AI.
The music created by AI is to put in a kind of signal sound, at frequencies that cannot be heard by the human ear. (Insert ultrasonic signal)
Well, I think this is a natural trend,
I also think it is an inevitable trend that there will eventually be a technology that eliminates this kind of marker (watermark) and allows AI generation without a marker. (Illegal act technology will be created)
Like there is legitimate music and pirated music,
Just like there is legitimate AI music and pirated AI music.
We live in an age where content is mass-produced by humans and AI,
“Is this a person or an AI?
I believe that the distinction and judgment will no longer be possible solely through the technology and management of the generating AI side (major AI companies).
So, what to do?
To stretch my imagination even further, I think,
I think the only way is for the person himself to prove whether the content is him (himself) or an AI.
In response to AI (fake) content,
The person in question refutes, “I didn’t say that, I didn’t sing that, I didn’t go there.”
So how do we prove the counter-argument?
It is to record 24/7 of your life, like a drive recorder.
That way, in one shot, we can tell if it is a fake video (AI content) or not, and we can prove it.
In such a future,
In order to eradicate fakes, technology and systems that record you 24/7 may become necessary in the future, and may be born.
This is the age of recording your own records to prove yourself.
Perhaps someday the time will come when we will long for the days when we could understand each other or be fooled just by saying, “It’s me, me, me,” over the phone, the analog and lax times.
Pardon my long-winded fantasy.
See you then.
Note: There was also this article.
YouTube Changes Policy. Cracking down on “look-alike” AI content]
After all, the control of AI content is left to the good will of the users, and there are limits, aren’t there?
T. Payne, one of the “U.S. Selected Creators,” was immediately using this feature to make music. It seems he is making music really easily, using his AI to create music to his liking (T. Payne’s liking). Thus, if you make full use of your own AI, there is no need for the above distinction and discrimination.^^;（It is only AI content created by others, fake content, that is the problem.）